On wind tunnels
A wind tunnel provides the experimental counterpart to CFD, measuring the aerodynamic forces on a physical model placed in a controlled airstream. Its role is to provide an independent point of validation for the numerical results. A closed-circuit tunnel is typically instrumented to measure the two integrated force coefficients of interest: downforce and drag.
The theoretical basis of all wind-tunnel testing is dynamic similarity: a scaled model reproduces the full-scale flow faithfully only if the relevant non-dimensional groups are matched between the two. For incompressible aerodynamics the governing group is the Reynolds number:
where \(c\) is a reference length, here the effective wing chord. Matching the Reynolds number for the same fluid (constant \(\mu\)) at incompressible regime (constant \(\rho\)) requires \(U_{\text{model}}\, c_{\text{model}} = U_{\text{real}}\, c_{\text{real}}\). Since a scaled model has a smaller chord, the tunnel speed would have to exceed the real speed to compensate, which for the very high speeds of a racing car is unattainable in a small tunnel. Full dynamic similarity is typically only attainable for F1 teams with tremendously large and fast wind tunnels, whereas their models are typically to 60% scale. A reduced-scale model in a modest tunnel can only reproduce the real flow at low speeds — a slow corner, say — so the Reynolds number does not match the modelled speed and correlation is not exactly held. This is a compromise forced by the constraints of the facility.
Because the model sits inside a bounded test section rather than in free air, its measurements are subject to systematic effects that must be kept small by sizing the model appropriately. Two rules of thumb govern this sizing:
- Blockage ratio. The model obstructs part of the tunnel cross-section, artificially accelerating the flow around it and inflating the measured forces. The effect is limited by keeping the model frontal area to a small fraction of the test-section area — conventionally only around 5 to 10% — beyond which correction factors become large and unreliable.
- Wall proximity. The proximity of the model to the tunnel walls distorts the flow field a lifting body would have in free air, an effect known as wall interference. It is contained by keeping a sufficient distance between the model and the walls, in practice by limiting the model span to a fraction of the tunnel width, so that the induced flow is not unduly constrained.
Standard correction procedures exist for the residual blockage and interference that remain once the model is sensibly sized.
A further consideration is specific to the reduced scale of the model. As noted above, the model cannot reach full-scale Reynolds numbers, and at low Reynolds number a significant portion of the boundary layer may remain laminar, whereas at full scale it would be turbulent almost everywhere. This mismatch matters because a laminar boundary layer separates far more readily than a turbulent one and would misrepresent the aerodynamics; moreover, the CFD against which the model is compared assumes a fully turbulent boundary layer. To reconcile the two, the boundary layer on the model is deliberately tripped — transition to turbulence is forced at a fixed location by a small surface disturbance, in this case, tape near the leading edge — so that the model, the full-scale wing and the simulation all share a turbulent boundary layer. Tripping does not recover the missing Reynolds number, but it ensures that the flow regime is representative, which is what makes the comparison meaningful.